Inspiration
Political language carries weight : a speech can steady a nation, a single sentence online can fracture it. Words move fast, but their impact is often invisible. We wanted to explore that invisible space, the emotional currents, the appeals to logic, the claims to credibility and bring classical rhetoric into dialogue with machine learning. Our goal : to make persuasion visible.
What it does
Speech Impact Analyzer takes any political text, a speech excerpt, a tweet, a statement, and uncovers the rhetorical forces inside it. It identifies ethos, pathos, and logos ; analyzes emotional tone and structure ; and predicts how strongly the message is likely to engage the public.
How we built it
We gathered speeches and tweets from political figures, then unified them into a single, coherent dataset. We cleaned each text carefully, removing noise while preserving meaningful punctuation, pronouns, and intensity markers. We engineered rhetorical features, built sentiment and linguistic indicators, and tested multiple ML models.
Challenges
Tweets are messy and impulsive while speeches are polished and deliberate. Engagement varies wildly depending on the speaker, so we had to normalize it to capture true rhetorical impact. The greatest challenge was quantifying something as human as persuasion without flattening its nuance.
What we learned
That rhetoric leaves fingerprints : subtle, persistent, measurable. And that combining ML with humanities can reveal patterns in political communication that human intuition alone often misses.
What’s next
We’d like to expand to multilingual political corpora, and map deeper rhetorical structures such as repetition, contrast, or metaphor.
Log in or sign up for Devpost to join the conversation.